Available — Project Name Finder

Always pass --json so you can parse and present results cleanly.

Generate names

available "description of the project" --json

Check specific names

available --check "name1,name2,name3" --json

Important: use defaults unless the user asks otherwise

The tool checks 3 TLDs (.com, .dev, .io) and 10 package registries by default. These defaults give the best results — do NOT pass --tlds, --registries, or --models unless the user specifically asks to customize them. Passing these flags limits the check to only what you specify, which produces worse results.

Optional flags (only if the user requests):

• --max-names 30 — generate more candidates (default: 20)
• --tlds com,dev,io,org — override which TLDs to check
• --registries npm,crates.io — override which registries to check (limits to ONLY these)
• --models claude-opus-4-6 — pick specific LLM models

Parsing the JSON Output

The tool outputs structured JSON to stdout (progress/warnings go to stderr):

{
  "results": [
    {
      "name": "rushq",
      "score": 0.8,
      "suggested_by": ["claude-opus-4-6"],
      "domains": {
        "available": 3,
        "registered": 0,
        "unknown": 0,
        "total": 3,
        "details": [
          {"domain": "rushq.com", "available": "available"},
          {"domain": "rushq.dev", "available": "available"},
          {"domain": "rushq.io", "available": "available"}
        ]
      },
      "packages": {
        "available": 9,
        "taken": 1,
        "unknown": 0,
        "total": 10,
        "details": [
          {"registry": "crates.io", "available": "available"},
          {"registry": "npm", "available": "taken"}
        ]
      }
    }
  ],
  "models_used": ["claude-opus-4-6"],
  "errors": []
}

Key fields:

• score: 0.0–1.0 availability score (higher is better). Weights: .com = 30%, other TLDs = 10% each, package registries = 50% split evenly.
• domains.details[].available: "available", "registered", or "unknown"
• packages.details[].available: "available", "taken", or "unknown"
• errors: Per-model generation failures (only in generate mode)

Presenting Results

After running the command and parsing JSON, present results to the user in a clear summary. Focus on the top candidates (score >= 0.7) and highlight what's actually available vs taken.

Formatting guide

For each top result, show:

1. The name and its score as a percentage
2. Domain availability — which TLDs are free vs taken
3. Package registry availability — summary count plus any notable conflicts (e.g., npm taken)
4. Which models suggested it (in generate mode)

Group results into tiers:

• Fully available (score >= 0.9): all or nearly all domains and registries clear
• Mostly available (score 0.7–0.9): some conflicts but still viable
• Partially available (score < 0.7): mention briefly, note key conflicts

If the user asked about specific names, give a direct answer for each: available or not, and where the conflicts are.

Errors

If errors is non-empty, mention which models failed but don't make it the focus — the results from other models are still valid. If all models fail in generate mode, tell the user to check their API keys.

Tips

• The tool needs at least one LLM API key set (ANTHROPIC_API_KEY, OPENAI_API_KEY, GOOGLE_API_KEY, or XAI_API_KEY) for generate mode. Check mode works without any keys.
• If the user wants to check a lot of names, use --check rather than generating — it's faster and doesn't need API keys.
• Results are sorted by score (best first), so the top entries are always the strongest candidates.
• If the binary isn't found, suggest: cargo install --git https://github.com/brad/available (adjust URL as needed).